As part of the New York High-Risk Project and in the context of a third round of testing, auditory short-term recognition memory for words and for consonant-vowel-consonant trigrams was measured in normal control adolescents (n = 53) and in adolescents at high risk for schizophrenia (n = 46). Differences in performance between the two groups are attributable to a lower initial memory strength on the part of the high-risk subjects, with trigrams showing larger differences than words. A subgroup of the high-risk subjects characterized by "clinical deviancy" showed, in addition, a (nonsignificant) increase in the rate of information loss from memory for trigrams.